Skip to main content

Table 1 (abstract P7). Multivariable models (before and after variable selection) for predicting spontaneous preterm delivery within seven days in symptomatic women

From: Methods for Evaluation of medical prediction Models, Tests And Biomarkers (MEMTAB) 2018 Symposium

 

Model including all variables

Model after variable selection

Beta

OR (95% CI)

Betaa

OR (95% CI)

Intercept

 Study 1

-7.9

 

-4.6

 

 Study 2

-8.5

 

-5.3

 

 Study 3

-9.0

 

-5.7

 

 Study 4

-8.7

 

-5.4

 

 Study 5

-9.3

 

-6.0

 

Quantitative fetal fibronectin (qfFN)b

 (qfFN+1)/100)^0.5

2.0

7.6 (5.7 – 10)

1.89

6.6 (4.9 – 8.9)

Age (yr)

0.02

1.0 (0.98 – 1.1)

-

-

BMI (kg/m2)

0.02

1.0 (0.96 – 1.1)

-

-

Smoking

-0.66

0.52 (0.24 – 1.1)

-0.67

0.51 (0.24 – 1.08)

Ethnicity

 1 Caucasian

Reference

 

Reference

 

 2 South Asian

1.1

2.9 (0.93 – 9.1)

0.94

2.6 (0.84 – 7.9)

 3 East Asian

-1.2

0.31 (0.04 – 2.5)

-1.004

0.37 (0.05 – 2.8)

 4 African, Caribbean, Middle-East

-0.21

0.81 (0.42 – 1.5)

-0.21

0.81 (0.43 – 1.5)

 5 Other

-0.25

0.78 (0.20 – 3.0)

-0.31

0.74 (0.19 – 2.8)

Nulliparity

0.53

1.7 (1.1 – 2.7)

0.37

1.4 (0.92 – 2.2)

Multiple pregnancy

0.85

2.3 (1.4 – 4.1)

0.83

2.3 (1.3 – 3.9)

Previous spontaneous

PTD < 34 weeks

0.43

1.5 (0.78 – 1.1)

-

-

Gestational age at assessment (wks)

0.03

1.0 (0.96 – 1.1)

-

-

Nagelkerke R2

0.39

0.39

AUC (95%CI)

0.90 (95% CI 0.88 – 0.93)

0.89 (95% CI 0.87 – 0.93)

  1. PTD preterm delivery < 7 days
  2. aRegression coefficients of the predictors in the model were shrunken with a uniform shrinkage factor 0.92. The intercepts were re-estimated after shrinkage of the regression coefficients of the predictors to ensure perfect calibration-in-the-large.
  3. bTransformation of continuous variable ‘quantitative fetal fibronectin’ because of non-linearity